Top 500 trends

Michael Huntingdon hunting at
Mon Nov 25 18:31:30 PST 2002

One might be quite surprised at the cost actually paid, when the resultant
systems may well end up high on the top 500 list. And in these cases, it's
those who must haul the heaviest of loads that purchase a truck rather than
station wagon. But it's far more than just the high end number crunching
ability of the solution. 

It's comes down to ensuring that every bit of data is without question the
one or the zero it's supposed to be. One missed bit, whether because of
less expensive memory, or the inability of the operating system to
checkpoint the data (100% of the time) at the system/option level...could
mean months and months of testing washed, or an assumption based on the
output that is totally incorrect. When assumptions about nuclear testing
are involved, I'd prefer my country spend whatever amount the true experts
feel is correct. Having worked with researchers for many years...I know two
things with certainty; they are the experts, and they DO NOT spend money
when or where it's not required. bet!

And I'm well aware that no system is hack-proof...but...certain operating
systems lend themselves to tighter security. In the case of the systems we
find at the top of the top 500, they simply fit the requirements of the
project. My hat is off to those who provided the initial engineering and
follow on work within the Quadrics environment. It will eventually trickle
down into the realm of products which might be considered more wolf-ish.

It takes vision, talent and money to breed new technology. In some cases it
also takes a purchase that hits the top 500 in order to see movement in one
direction or another. This too is factored in when projects like this are
put into place. 

A great deal of emerging technology also comes from our universities. But
again, even in these situations, it typically doesn't come without the
financial support of an organization outside a campus. During times when
everyone is cutting back, how about just a little applause every now and
then for those with the courage to put their money where their mouth is,
like those who developed Quadrics.

I'm surprised to see anyone with background and interests in ASCI or ROCKS
who is not more sensitive to the roots of TCO. Or perhaps I misunderstood
the direction being taken in this stream.


At 08:31 PM 11/25/2002 -0500, you wrote:
>On Mon, 2002-11-25 at 19:49, Michael Huntingdon wrote:
>> Pure and simple, there are those applications and computing environments
>> that require more. One size does not fit all.
>Of course one size machine, architecture, budget does not fit all.  That
>much is obvious.
>The question I asked was what the actual costs are involved in these
>"best efforts".  Moreover I posited that they might belie the claims of
>"low costs" of certain architectures, or TCO's of others.  I further
>posited that these TCOs may not render such machines viable from the
>purchasers point of view.
>Joseph Landman, Ph.D
>Scalable Informatics LLC
>email: landman at
>  web:
>phone: +1 734 612 4615

More information about the Beowulf mailing list